I have difficulty cold-resetting Tyan s2912, which has mcp55.
The sequence outb(0x2,0xcf9); outb(0x6,0xcf9); is a warm reset. It does not reset the HyperTransport link frequency and width.
The sequence outb(0xa,0xcf9); outb(0xe,0xcf9); causes the CPU to hang but does not reset the MB.
Does anybody know what is wrong?
Thanks, Roman
On Feb 4, 2008 9:29 PM, Roman Kononov kononov@dls.net wrote:
I have difficulty cold-resetting Tyan s2912, which has mcp55.
The sequence outb(0x2,0xcf9); outb(0x6,0xcf9); is a warm reset. It does not reset the HyperTransport link frequency and width.
how do you know?
do you have other co-processor or htx installed?
YH
On 2008-02-05 00:48 yhlu said the following:
On Feb 4, 2008 9:29 PM, Roman Kononov kononov@dls.net wrote:
I have difficulty cold-resetting Tyan s2912, which has mcp55.
The sequence outb(0x2,0xcf9); outb(0x6,0xcf9); is a warm reset. It does not reset the HyperTransport link frequency and width.
how do you know?
do you have other co-processor or htx installed?
I have a co-processor.
The sequence I hope to have:
1) Power up. Power-on reset. The co-processor takes long time to become alive (~500s). By this time the CPU thinks that the HT link is dead. 2) CPU issues another reset. 3) The CPU re-initializes the HT link and the connection is happy.
With the above sequence, the step 3) is a warm reset, and the CPU does not attempt to re-initialize the link.
YH
Roman
On 2008-02-05 09:48 Roman Kononov said the following:
On 2008-02-05 00:48 yhlu said the following:
On Feb 4, 2008 9:29 PM, Roman Kononov kononov@dls.net wrote:
I have difficulty cold-resetting Tyan s2912, which has mcp55.
The sequence outb(0x2,0xcf9); outb(0x6,0xcf9); is a warm reset. It does not reset the HyperTransport link frequency and width.
how do you know?
do you have other co-processor or htx installed?
I have a co-processor.
The sequence I hope to have:
- Power up. Power-on reset. The co-processor takes long time to become
alive (~500s). By this time the CPU thinks that the HT link is dead.
~0.5s here.
- CPU issues another reset.
- The CPU re-initializes the HT link and the connection is happy.
With the above sequence, the step 3) is a warm reset, and the CPU does not attempt to re-initialize the link.
YH
Roman
On Feb 5, 2008 7:48 AM, Roman Kononov kononov@dls.net wrote:
On 2008-02-05 00:48 yhlu said the following:
On Feb 4, 2008 9:29 PM, Roman Kononov kononov@dls.net wrote:
I have difficulty cold-resetting Tyan s2912, which has mcp55.
The sequence outb(0x2,0xcf9); outb(0x6,0xcf9); is a warm reset. It does not reset the HyperTransport link frequency and width.
how do you know?
do you have other co-processor or htx installed?
I have a co-processor.
The sequence I hope to have:
- Power up. Power-on reset. The co-processor takes long time to become
alive (~500s). By this time the CPU thinks that the HT link is dead. 2) CPU issues another reset. 3) The CPU re-initializes the HT link and the connection is happy.
With the above sequence, the step 3) is a warm reset, and the CPU does not attempt to re-initialize the link.
that is co-processor problem.
Please talk the co-processor vendor about that. they have solution for you problem. otherwise return that to get one from another vendor.
YH
On 2008-02-05 11:45, yhlu wrote:
On Feb 5, 2008 7:48 AM, Roman Kononov kononov@dls.net wrote:
On 2008-02-05 00:48 yhlu said the following:
On Feb 4, 2008 9:29 PM, Roman Kononov kononov@dls.net wrote:
I have difficulty cold-resetting Tyan s2912, which has mcp55.
The sequence outb(0x2,0xcf9); outb(0x6,0xcf9); is a warm reset. It does not reset the HyperTransport link frequency and width.
how do you know?
do you have other co-processor or htx installed?
I have a co-processor.
The sequence I hope to have:
- Power up. Power-on reset. The co-processor takes long time to become
alive (~500s). By this time the CPU thinks that the HT link is dead. 2) CPU issues another reset. 3) The CPU re-initializes the HT link and the connection is happy.
With the above sequence, the step 3) is a warm reset, and the CPU does not attempt to re-initialize the link.
that is co-processor problem.
Please talk the co-processor vendor about that. they have solution for you problem. otherwise return that to get one from another vendor.
Thanks for the recommendation. Who makes socket F co-processors?
Can you tell me how to make the cold reset please?
Roman
On 2008-02-05 11:45, yhlu wrote:
On Feb 5, 2008 7:48 AM, Roman Kononov kononov@dls.net wrote:
On 2008-02-05 00:48 yhlu said the following:
On Feb 4, 2008 9:29 PM, Roman Kononov kononov@dls.net wrote:
I have difficulty cold-resetting Tyan s2912, which has mcp55.
The sequence outb(0x2,0xcf9); outb(0x6,0xcf9); is a warm reset. It does not reset the HyperTransport link frequency
and width.
how do you know?
do you have other co-processor or htx installed?
I have a co-processor.
The sequence I hope to have:
- Power up. Power-on reset. The co-processor takes long time to become
alive (~500s). By this time the CPU thinks that the HT link is dead. 2) CPU issues another reset. 3) The CPU re-initializes the HT link and the connection is happy.
With the above sequence, the step 3) is a warm reset, and the CPU does not attempt to re-initialize the link.
that is co-processor problem.
Please talk the co-processor vendor about that. they have solution for you problem. otherwise return that to get one from another vendor.
Thanks for the recommendation. Who makes socket F co-processors?
Can you tell me how to make the cold reset please?
Isn't there supposed to be a way to use ldtstop to reinitialize the HT link without a cold reset? It seems like I read that it is faster.
Myles
On 2008-02-05 13:29, Myles Watson wrote:
Isn't there supposed to be a way to use ldtstop to reinitialize the HT link without a cold reset? It seems like I read that it is faster.
No, ldtstop does not fully reinitialize the link. If the CPU detects a dead link after the cold reset, it will never reinitialize that link. This comes from the experiments.
Roman
On Feb 5, 2008 11:13 AM, Roman Kononov kononov@dls.net wrote:
On 2008-02-05 11:45, yhlu wrote:
On Feb 5, 2008 7:48 AM, Roman Kononov kononov@dls.net wrote:
On 2008-02-05 00:48 yhlu said the following:
On Feb 4, 2008 9:29 PM, Roman Kononov kononov@dls.net wrote:
I have difficulty cold-resetting Tyan s2912, which has mcp55.
The sequence outb(0x2,0xcf9); outb(0x6,0xcf9); is a warm reset. It does not reset the HyperTransport link frequency and width.
how do you know?
do you have other co-processor or htx installed?
I have a co-processor.
The sequence I hope to have:
- Power up. Power-on reset. The co-processor takes long time to become
alive (~500s). By this time the CPU thinks that the HT link is dead. 2) CPU issues another reset. 3) The CPU re-initializes the HT link and the connection is happy.
With the above sequence, the step 3) is a warm reset, and the CPU does not attempt to re-initialize the link.
that is co-processor problem.
Please talk the co-processor vendor about that. they have solution for you problem. otherwise return that to get one from another vendor.
Thanks for the recommendation. Who makes socket F co-processors?
I don't think there are hundreds of vendors for that.
DRC computer?
Can you tell me how to make the cold reset please?
you don't need cold reset.
at that case even cold reset doesn't work.
YH
On 2008-02-05 14:18, yhlu wrote:
I don't think there are hundreds of vendors for that.
DRC computer?
They offer only socket 940.
you don't need cold reset.
at that case even cold reset doesn't work.
After "HT Cold Reset" (PWROK deasserted and SYSRST asserted) the CPU reinitializes its links. This is proven.
Roman.
On 2008-02-05 09:48, Roman Kononov wrote:
I have a co-processor.
The sequence I hope to have:
- Power up. Power-on reset. The co-processor takes long time to become
alive (~500s). By this time the CPU thinks that the HT link is dead. 2) CPU issues another reset. 3) The CPU re-initializes the HT link and the connection is happy.
With the above sequence, the step 3) is a warm reset, and the CPU does not attempt to re-initialize the link.
Never mind. I found the way to make the reset long enough.
Thanks,
Roman
On Feb 5, 2008 3:29 PM, Roman Kononov kononov@dls.net wrote:
On 2008-02-05 09:48, Roman Kononov wrote:
I have a co-processor.
The sequence I hope to have:
- Power up. Power-on reset. The co-processor takes long time to become
alive (~500s). By this time the CPU thinks that the HT link is dead. 2) CPU issues another reset. 3) The CPU re-initializes the HT link and the connection is happy.
With the above sequence, the step 3) is a warm reset, and the CPU does not attempt to re-initialize the link.
Never mind. I found the way to make the reset long enough.
then cold reset and warm reset all work?
YH
On 2008-02-05 19:06, yhlu wrote:
then cold reset and warm reset all work?
These things are still the same. Warm reset works. Cold reset does not. After the cold reset sequence the CPU hangs, only the push button or power switch can revive it. Is it supposed to work or there are secrets somewhere?
Roman
On Feb 5, 2008 5:19 PM, Roman Kononov kononov@dls.net wrote:
On 2008-02-05 19:06, yhlu wrote:
then cold reset and warm reset all work?
These things are still the same. Warm reset works. Cold reset does not. After the cold reset sequence the CPU hangs, only the push button or power switch can revive it. Is it supposed to work or there are secrets somewhere?
if replace the co-processor with opteron, does the cold reset work?
YH
On 2008-02-05 19:41, yhlu wrote:
if replace the co-processor with opteron, does the cold reset work?
I did not try 2 processors. With a single processor alone it does not work.
Roman
-
Myles Watson -
Roman Kononov -
yhlu